Skip to main content
Risk-Management Case Studies

Societe Generale and Jerome Kerviel

Pomegra Learn

How Did Jerome Kerviel Hide €50 Billion in Unauthorized Positions?

On January 24, 2008, Societe Generale announced that one of its traders, Jerome Kerviel, had accumulated approximately €50 billion ($65 billion) in unauthorized positions—bets that had never been approved, disclosed to management, or tracked in the firm's risk management systems. When discovered, the positions had deteriorated badly, and Societe Generale was forced to liquidate all of Kerviel's trades in a single day, realizing a loss of €4.9 billion ($6.3 billion), one of the largest single-trader losses in financial history.

What made Kerviel's fraud remarkable was not the size of the positions (though €50 billion was enormous) but the fact that a junior derivatives trader with a base salary of €120,000 per year was able to hide these positions for months using fraudulent documentation and exploiting gaps in the bank's control procedures. Kerviel did not steal money or hack systems; he simply violated position limits using procedures that the bank's risk management framework should have caught.

Quick definition: Rogue trader risk refers to the potential for loss caused by a trader or employee who executes unauthorized or fraudulent trades, frequently by falsifying documentation, impersonating supervisors, or exploiting control gaps to bypass position limits and risk oversight.

Key takeaways​

  • Kerviel used fictitious hedging trades (fake sales) to offset his massive long positions in stock index futures, allowing the real positions to remain hidden from the bank's VaR and position limit calculations
  • The bank's compliance systems were segregated in ways that allowed Kerviel to see both sides of transactions but prevented compliance officers from seeing the complete picture of his positions
  • Back-office reconciliation procedures (supposed to catch discrepancies between trading and settlement records) failed to identify the fake hedges, enabling the fraud to persist for months
  • Risk systems calculated VaR and position limits based on net positions after offsets, but Kerviel's fake hedges inflated the apparent offset, making the real risk invisible
  • Societe Generale's supervisory structure allowed Kerviel to report to a supervisor he had once worked with, creating a social relationship that may have hindered objective risk monitoring

The Setup: How a Junior Trader Became a Risk Manager's Nightmare​

Jerome Kerviel was born in 1977 and began working at Societe Generale in 2000 as a junior analyst. By 2005, he had been promoted to equity derivatives trader. In this role, Kerviel was responsible for executing client trades in equity index futures and equity-linked derivatives. His typical job was to facilitate client orders and manage small proprietary positions.

Kerviel's compensation structure aligned with the bank's implicit incentive to take on risk: he earned bonuses proportional to the profits he generated. In the trading environment of 2005–2007, larger positions and higher leverage generated higher returns, and therefore higher bonuses. This created a powerful incentive for Kerviel to grow his book.

The critical vulnerability in Societe Generale's control structure was that Kerviel occupied a unique position: he was both a trader and an expert in the back-office reconciliation procedures. Because he had worked in the back-office earlier in his career, Kerviel understood exactly how the bank's systems worked, where the gaps were, and how to exploit them.

Kerviel's strategy was straightforward: he would accumulate massive long positions in DAX (German stock index) futures and other equity index products. These positions were legitimate—he was authorized to trade them. But the size of his positions vastly exceeded the limits he was authorized to hold. To hide the positions from the risk management system, Kerviel created fictitious offsetting hedges.

Specifically, he would:

  1. Execute a genuine long position (buy €1 billion in DAX index futures)
  2. Immediately create a fictitious sale trade (sell €1 billion in the same index, but with a fake counterparty)
  3. Record both trades in the system
  4. The net position would appear to be zero from the risk management system's perspective
  5. The real long position would remain hidden

The fake offsetting sale would be attributed to a purported hedge or client trade, but the counterparty was fictitious, and no actual transaction occurred. When settlement time came (a few days after the trade), Kerviel would falsify the settlement documentation to make it look like the fake trade had settled normally.

The Cascade: Why Controls Failed at Multiple Levels​

Societe Generale's control framework had several layers designed to catch unauthorized traders:

Position Limit Monitoring: Traders had authorized position limits (e.g., maximum €50 million in DAX futures). Kerviel was authorized for small positions, perhaps €20–50 million. His actual position reached €50 billion. The system should have flagged this immediately. It didn't—because the fake offsets made the net position appear acceptable.

Value-at-Risk (VaR) Calculation: Risk management calculates VaR on net positions to ensure traders aren't taking on uncompensated risk. Kerviel's VaR should have spiked as his positions grew. It didn't—because the system used the offsetting positions to calculate VaR, which appeared small.

Back-Office Reconciliation: The back-office is responsible for reconciling trades recorded in the trading system with actual settlements and confirmations. This is where fake trades should have been caught. The back-office staff should have received confirmations from counterparties saying "we have no record of this trade," which would have flagged Kerviel's fraud. Kerviel exploited the timing gap: trades settle in T+2 or T+3 (2–3 days after execution). During this window, fake trades sat in the system unconfirmed. Kerviel would often cancel or reverse them before settlement, making them disappear before the back-office could receive counterparty confirmations.

Segregation of Duties: In a well-designed control environment, the trader, the supervisor, the back-office, and the risk management team should be independent. If one person controls both the trade and the documentation, fraud becomes easier. Kerviel had access to trading systems, settlement procedures, and internal documentation. This combination should have been prevented.

Supervisor Review: Each trader reports to a supervisor who is responsible for monitoring their activities. Kerviel's supervisor, Olivier Deville, was someone Kerviel had worked with previously. Deville may not have exercised appropriate skepticism or may have been swayed by Kerviel's personal relationship and apparent trading success.

How Kerviel's Fraud Worked​

This diagram shows how Kerviel exploited the timing between trade execution and settlement confirmation to hide his positions.

The Fraud Mechanics: Staying Invisible​

For Kerviel's fraud to persist from early 2007 through January 2008, he needed to continually refresh and hide his offsetting trades. As the real position grew larger, the fake offsets also grew. By January 2008, Kerviel had:

  • Approximately €50 billion in genuine long positions in equity index futures
  • Approximately €50 billion in fictitious offsetting positions created through fake trades
  • A net position that appeared near zero to the risk management system

The timing of discovery was driven by external market conditions. In mid-January 2008, as markets declined in response to U.S. credit concerns, the long equity positions that Kerviel held began losing value. On January 18, 2008, markets fell sharply. Kerviel's positions deteriorated by approximately €1 billion. He realized he needed to hide the losses, and he attempted to execute additional fake trades to offset the real losses.

However, on January 21, 2008, a back-office employee noticed a discrepancy: a trade in Kerviel's name did not match settlement instructions. This triggered an investigation. Within days, the full scope of Kerviel's fraud became apparent. On January 24, 2008, Societe Generale announced the discovery.

Why Supervisory Controls Failed​

Olivier Deville, Kerviel's direct supervisor, had approved many of Kerviel's trades. In a well-designed control environment, a supervisor should verify that trades match authorized limits and are appropriately documented. Deville did not. Several factors contributed to this failure:

Personal Relationship: Deville and Kerviel had worked together previously. This relationship may have created a bias toward trust. Supervisors should be taught to maintain professional distance, but in practice, personal relationships influence judgment.

Performance Bias: Kerviel had generated substantial trading profits (on paper) in 2006 and 2007. His apparent success may have made Deville less inclined to question his activities. This is a common bias in risk management: successful traders get more leeway.

Communication Gaps: Deville may not have been fully informed about Kerviel's actual position sizes. The risk management systems showed net positions, not gross positions. Deville may have believed the net positions were actually small and that Kerviel's trading was within bounds.

Lack of Skepticism Training: Trading supervision often assumes that traders are trying to maximize profits within authorized bounds. There's less emphasis on the possibility that traders might simply lie. Supervisors should be trained to maintain healthy skepticism and to spot red flags like traders reversing trades with unusual frequency.

Real-world examples​

Nick Leeson and Barings Bank (1995): Nick Leeson was a derivatives trader at Barings Bank who accumulated approximately £860 million in unauthorized positions through fictitious offsetting trades. Like Kerviel, Leeson had access to both front-office and back-office systems. When discovered, the positions had deteriorated so badly that Barings Bank, founded in 1762, became insolvent. Barings went bankrupt in February 1995, and Leeson fled to Malaysia before being extradited and prosecuted.

Kweku Adoboli at UBS (2011): Three years after Kerviel's fraud, UBS announced that trader Kweku Adoboli had hidden approximately $2 billion in unauthorized positions through similar techniques—creating fake hedging trades that inflated the apparent offset while hiding real risks. Adoboli's fraud cost UBS an estimated $2.3 billion in losses and legal settlements.

Akif Khan at BNP Paribas (2002): A junior trader at BNP Paribas accumulated €6.4 billion in unauthorized positions through fictitious trades. The pattern was identical: real long positions, fake offsetting positions, hiding the fraud through documentation gaps and timing mismatches.

Jérôme Kerviel's Conviction (2010): Kerviel was convicted in 2010 and sentenced to three years in prison (suspended), a fine of €300,000, and required to pay Societe Generale €4.9 billion in restitution. His conviction was later partially overturned, but the core fraud was never disputed.

Common mistakes​

1. Assuming that traders have incentive to behave honestly because they're well-compensated Kerviel earned approximately €300,000–400,000 per year (base salary plus bonus in 2006–2007), which was substantial but not exceptional compared to senior managing directors or stars of the firm. His compensation was tied to profits, which created an incentive to grow his book and take on risk. The implicit message was "generate profits, and we'll reward you." Kerviel's fraud could be seen as an extreme response to this incentive structure.

2. Allowing the same person to have responsibility for both trading and back-office reconciliation Kerviel had worked in the back-office and understood exactly how reconciliation procedures worked. He could exploit timing gaps and create fake trades that would disappear before they were confirmed by counterparties. Segregation of duties should prevent any trader from having detailed knowledge of how to circumvent back-office controls.

3. Using net position reporting for risk management when gross positions would reveal the fraud The risk management system reported Kerviel's net position (long position + fake offset ≈ zero). If the system had also reported gross notional exposure (total value of all positions, regardless of offset), the fraud would have been apparent. Supervisors should see both net and gross exposure.

4. Assuming that sophisticated risk systems are sufficient without human oversight Societe Generale had Value-at-Risk models, position limit monitoring, and back-office reconciliation. All of these systems "worked" in the sense that they did what they were designed to do. However, they didn't work together, and there was a gap in human judgment. A supervisor who actually reviewed trading confirmations in detail would have noticed the fake trades immediately.

5. Not training supervisors to spot behavioral red flags Kerviel reversed trades with unusual frequency (to remove evidence of the fake offsets before settlement). He made unusual bookkeeping entries and adjustments. He requested settlement delays and accommodations. These are behavioral red flags that a well-trained supervisor should catch. Instead, supervisors were focused on performance metrics and market conditions.

6. Failing to implement a canary control A simple procedure would have been for the back-office to send periodic confirmations to external counterparties: "We have a record of a trade with you on this date and amount. Please confirm." Any fake trade with a nonexistent counterparty would have been caught immediately. Societe Generale didn't implement this control until after Kerviel's fraud was discovered.

FAQ​

How did Kerviel become a trader if he was a back-office employee?​

Kerviel started at Societe Generale as a junior analyst in the compliance/back-office area around 2000. By 2005, the bank had promoted him to derivatives trader. This promotion shows both an opportunity (he understood the systems) and a failure of judgment by the bank (he understood exactly how to exploit the systems).

Why was Kerviel's fraud so easy to hide if the bank had sophisticated risk systems?​

The systems were "sophisticated" in the sense that they were technologically advanced, but they had logical gaps. The systems calculated risk on net positions, not gross positions. They didn't cross-check trading records with external counterparty confirmations (or did so with long delays). They allowed timing gaps between trade execution and settlement where fake trades could be reversed before being caught.

What would have prevented Kerviel's fraud?​

Several layers of control would have prevented it: (1) gross position reporting (supervisors would see the large notional exposure), (2) external counterparty confirmation (any fake trade would be caught immediately), (3) segregation of duties (Kerviel shouldn't have knowledge of back-office timing gaps), (4) regular trading audits by independent compliance staff, and (5) systems that automatically reject trades that exceed position limits without specific authorization from senior management.

Did Societe Generale's insurance cover the losses?​

No. Societe Generale's insurance and risk management controls should have covered the loss, but the fraud was so large and clear-cut that insurance exclusions likely applied. Societe Generale bore the full €4.9 billion loss. The firm later settled with regulators by paying fines related to inadequate controls.

Could this type of fraud happen at a modern bank?​

Potentially yes, though most banks have strengthened controls since 2008. Key changes include: (1) gross position reporting, (2) automated counterparty confirmation procedures, (3) limits on trader access to back-office systems, (4) segregation of supervisory responsibilities, and (5) regulatory oversight of trader limits. However, trading fraud fundamentally depends on humans exploiting control gaps, so the risk never goes to zero.

Was Jerome Kerviel required to repay the €4.9 billion?​

Legally, yes. However, given Kerviel's limited personal wealth (he earned perhaps €300,000–400,000 per year), the likelihood of actually collecting €4.9 billion is near zero. Societe Generale pursued the legal judgment, but it's largely symbolic. Kerviel did serve a suspended prison sentence and faced legal consequences, but the practical recovery of damages was minimal.

How much should traders be allowed to lose before fraud is assumed?​

There's no bright-line answer, but the pattern matters more than the absolute number. A trader who makes a large bet, loses money, and admits the loss (even if it's billions) is different from a trader who hides positions, creates fake documentation, and falsifies records. The latter pattern is fraud; the former is risk management failure. Modern banks look at not just the magnitude of losses but whether the positions were authorized and accurately reported.

Summary​

Jerome Kerviel's fraud at Societe Generale in 2007–2008 represents a rogue trader case where a junior derivatives trader accumulated approximately €50 billion in unauthorized positions by creating fictitious offsetting trades and exploiting timing gaps in the bank's settlement and reconciliation procedures. Kerviel's net position appeared acceptable to risk systems, but his gross exposure was enormous. When markets moved against the real positions in January 2008, the fraud was discovered, and Societe Generale was forced to liquidate all positions in a single day, realizing a €4.9 billion loss.

The key lessons are: (1) risk systems should monitor gross positions and gross notional exposure, not just net positions, (2) back-office procedures must include external counterparty confirmation, (3) supervisors must review both trading records and settlement documentation, (4) segregation of duties prevents traders from knowing how to exploit control gaps, and (5) behavioral red flags (frequent trade reversals, documentation anomalies) should be monitored by supervisors. Modern banks have strengthened controls since Kerviel's fraud, but rogue trader risk remains a concern in any organization where individuals have discretion to execute trades and control documentation.

Next​

→ Bernie Madoff and the Failure of Due Diligence